Increasingly, online publication is replacing conventional, physical means for distributing content. For example, newspaper publishers are receiving increasing volumes of traffic at their websites, while selling fewer hard copies of the newspapers themselves. This presents a challenge for the publishers in terms of monetising the content they produce. While some publishers have instigated a pay-per-view or subscription service to access their online content, most have attempted to keep their online offerings free at the point of use. To support this model, publishers rely on advertising.
Indeed, advertising based revenue models are not limited to publishers of content but have been adopted by a large number of internet services. In selling advertisement space on a website, a decision has to be made as to the price that that space will cost. Today, advertising space is predominantly sold on either a pay-per-impression or a pay-per-click basis.
In a pay-per-impression model, a charge is levied each time the web page containing the advertisement is viewed. Records are kept each time a web page containing the advertisement is downloaded to a user, and a charge is levied to the advertiser according to the volume of downloads that have occurred.
A problem with current pay-per-impression techniques is that they do not accurately reflect whether the user engaged with the advertisement itself. Indeed, there is no indication as to whether the advertisement was even viewed by the user. For example, the advertisement may be on a part of the web page that was not viewable in the user's browser window, or may even have failed to load. This creates uncertainty as to the value of the advertisement to the advertiser, and thereby suppresses demand for internet advertising.
The pay-per-click model addresses some of these concerns. In this model, payment is made for each occasion on which the user clicks on a link within the advertisement. However, while this does ensure (in the absence of fraud) that payment is only made in the case where a user engages with the advertisement, it is not appropriate for all circumstances. For example, many advertisements are intended to increase brand awareness passively, and would not wish to interfere with the user's experience. In general, the number of clicks on an advertisement will not be simply a function of the number of views it receives, but will be affected by the success of that advertisement in enticing a click. Pay-per-click models are unable to distinguish between these factors.
Accordingly, it would be desirable to develop an approach which accurately reflects the number of times a particular advertisement has been viewed. Since the visibility of the advertisement is not ensured simply because the host web page has been downloaded, a greater degree of accuracy is required.
One proposed solution is to monitor the area of the web page which is shown by a particular user's browser at a given moment. For example, knowing the dimensions of the browser's viewport and the position of the advertisement in the web page, it can be established whether the advertisement is in view (that is, within the browser's viewport) by observing the scroll position of the browser window (i.e. its vertical and horizontal position on the web page). Another solution alternatively or additionally senses the position of the mouse cursor on the web page.
While these solutions can provide accurate data in the right circumstances, they are in practice of limited use. This is due to restrictions on the information that parties within the process are able to derive.
In particular, advertisements are often contained within (nested) iframes in a host web page. An iframe provides a region within a web page in which other web content may appear.
One important benefit of iframes for the publisher is that content within the iframe is unable to affect content in the host web page or even discover information regarding that page, as long as the iframe and the host web page are served from different domains. This means that the publisher can be secure in the fact that the advertisement will not corrupt the main content of the host web page.
However, since content within the iframe is not able to interact with the host web page, it is not possible for any client-side code served with the advertisement to discover, for example, the dimensions (height and width) of the host web page, or where the browser viewport is in relation to the host web page. Accordingly, it is not possible for any such client-side code to be used to discover whether the advertisement is in view. As such, when an iframe is used to serve an advertisement, techniques which attempt to infer whether an advertisement is in view from the location of the browser viewport relative to the host web page do not function. (c.f. http://www.realvu.net/iframes/)
Accordingly, it would be desirable to provide a technique for establishing whether an advertisement has been viewed, even when the advertisement is served within an iframe or some other such restricted or sandboxed environment.